Associations of Mental Health and Family Background With Opioid Analgesic Therapy

A Nationwide Swedish Register-Based Study

Patrick D. Quinn; Martin E. Rickert; Johan Franck; Amir Sariaslan; Katja Boersma; Paul Lichtenstein; Henrik Larsson; Brian M. D'Onofrio


Pain. 2019;160(11):2464-2472. 

In This Article


New initiation of noncancer opioid analgesic prescription was common in Sweden from 2007 onward. As in the United States,[75] receipt of LTOT was less common. An estimated 7.6% of individuals transitioned to a broader definition of LTOT—and 1.7% to a stricter definition—within 3 years of initiating an episode of opioid therapy. There were clear differences in rates of opioid receipt among those with and without preexisting mental health and family background factors. Individuals with diverse mental health conditions had greater rates of initiating opioid therapy and, especially, transitioning to LTOT. They also had greater rates of receiving benzodiazepines during opioid therapy, a deviation from current guidelines. Finally, opioid initiation and LTOT were also more common among those with parents with mental health conditions and those from lower SES families and neighborhoods.

The substantial associations of preexisting mental health conditions with LTOT are consistent with our previous results from US commercial health care insurance claims.[51,52] Other US studies have also found greater concurrent benzodiazepine and other sedative-hypnotic use among opioid recipients with mental health conditions.[63,80] This consistency is notable, given variability in populations, health care systems, and data sources. Indeed, there were striking differences in the composition of dispensed opioids. Although the weaker drugs codeine and tramadol comprised 65% of opioids in this study, the most commonly dispensed opioids in our US data were hydrocodone and oxycodone. Despite differences in potency, however, even weaker opioids may have abuse potential.[59]

There is also precedent for the current associations in previous Nordic-register-based research. Danish and Norwegian studies have found associations of, for example, smoking and benzodiazepine receipt with LTOT.[5,21] Although not all prior studies have found similar associations for substance use problems, inconsistencies may be due to differences in assessed behavior (eg, use vs SUD), measurement (eg, self-report vs clinical diagnosis), design (eg, surveys vs registers), and adjustment strategies.[21]

Although socioeconomic disparities in opioid receipt have been documented,[44,77] fewer studies have examined families of origin in this context.[39,57] As a result, there has been uncertainty regarding underlying processes and, in particular, the direction of the associations. For example, chronic pain—or LTOT itself—might disrupt socioeconomic advancement, thereby producing associations between opioid receipt and lower SES. By showing that lower SES in childhood predicted greater opioid receipt later in life, our results suggest that the association is not exclusively caused by an individual's own pain or opioid use leading to reduced SES. Rather, individuals from lower SES backgrounds may be more likely to receive opioid prescription, perhaps because of greater risk of occupational injury or limited access to nonpharmacologic treatments. However, the familial transmission of SES, pain, and mental health results from complex genetic and social pathways that must be elucidated to fully understand these associations.[34,68]

Taken together, the present results illustrate that the adverse selection pattern of opioid prescription among patients with mental health and other psychosocial problems is pervasive in Sweden. In light of the cross-national generality of this finding, comorbidity and shared (ie, transdiagnostic) liabilities in chronic pain and mental health may be more responsible for the prescription pattern than are factors specific to the US context, such as inappropriate or excessive opioid prescription.[6,30] Multiple explanations for the overrepresentation of depression, anxiety disorders, and SUDs in patients with chronic pain and LTOT have been proposed. There is evidence that mental health and chronic painful conditions partially share genetic, neural, and psychological mechanisms.[29,34,38] In particular, a recent review highlighted the shared processing of reward, pain, and social bonding and stress by the endogenous opioid system.[4] This work has given rise to the hypothesis that patients with depression may self-select into—and providers may be willing to prescribe—LTOT to not only manage pain but also self-medicate psychological distress.[71,73]

Given the breadth of mental health conditions associated with LTOT, as well as the stronger associations for SUD than for depressive or anxiety disorders, the extent to which emotion regulation or other motivations for opioid analgesic use are mental-health-condition-specific or more general remains to be determined.[69] At its epidemiologic level of analysis, this study cannot determine the processes that may have led to these prescription patterns. Opioid prescription decisions are proximally shaped by interpersonal interactions between patients and providers that would not be fully captured by health care–record data.[72] Further research with detailed clinical, genetic, or neurobiological assessment is needed to disentangle the pathways relating pain and LTOT to mental health comorbidity and other psychosocial problems.

Regardless of the specific processes, our findings provide further evidence that patients receiving opioids—and others with chronic noncancer pain—should receive thorough mental health assessment and, when appropriate, treatment.[14,43] The findings serve to underscore the clinical complexities and needs at the intersection of chronic pain, mental health comorbidity, and psychological distress.[31] Given that LTOT is often arrived upon unintentionally and can be difficult to taper for some,[20,30,35] mental health evaluation may help inform decisions to initiate or not initiate opioid therapy. Moreover, our findings add to evidence that opioid recipients with preexisting mental health conditions may be at subsequent risk of concurrent benzodiazepine therapy, a potentially harmful outcome[67] that may increase risk of opioid overdose.[76]

This study shares methodological advantages and disadvantages with other health care–record studies. First, we examined 8 years of nationwide opioid prescriptions, although we did not assess particular analgesia indications, including somatic pain locations or pain severity or sensitivity.[60] US studies have, however, found comparable mental health associations with and without restriction to diagnosed painful conditions.[26,51,52] Nevertheless, given the likely co-occurrence of chronic pain with the mental health and familial factors considered here, our results should be interpreted as describing differences in opioid prescription patterns rather than specific mental health contributions to a causal pathway to opioid initiation or LTOT. Second, our results concern opioid prescriptions written by providers and dispensed to patients. Records do not provide explicit information about primary or secondary noncompliance or illicitly obtained medications. We also could not determine whether providers used any risk-reduction strategies for opioid or concurrent benzodiazepine prescription. We note, though, that using a stricter LTOT criterion actually strengthened the mental health associations. Third, we examined opioid duration and strength but not dosage.[28] Fourth, we assessed diverse mental health conditions. Many registered diagnoses are validated,[41] although they would not capture patients who did not obtain specialist treatment (eg, those treated in primary care), which likely underestimated the prevalence of some conditions. However, differing predictor definitions—including psychoactive medications—produced comparable results. Fifth, we required at least 1.5 years' washout before the initiation of new episodes of opioid therapy.[62] Like previous health care-record–based opioid research,[5,9,24,51] however, this study cannot ensure that the cohort had never received opioids before the available data. Finally, we do not know the generalizability of our family-background findings to older individuals.


Understanding which patients select or are selected into opioid therapy and LTOT is necessary to assessing the real-world risk of iatrogenic SUD and other possible harms of opioid prescription. Indeed, recent reviews have underscored limitations in the literature on adverse opioid outcomes, notably including lack of adequate adjustment for confounders such as preexisting mental health conditions.[11,54] Register-based approaches can offer the generalizability, statistical power, and rich social, familial, and health data needed for rigorous analysis of opioid harms. At the same time, although they certainly do not rule out the possibility of such harms, our findings highlight the importance, in future research, of accounting for relevant preexisting differences between opioid recipients and nonrecipients beyond their own recorded medical history, such as SES and familial mental health liability.